DEGREE PROJECT, IN, SECOND LEVEL STOCKHOLM, SWEDEN 2015
Analysis of concussions in
professional ice hockey
JONAS SVENSSON
Sammanfattning
Huvudskador som hjärnskakning är ett problem i fullkontaktssporter som ishockey, dessa skador leder oftast till någon form av rehabilitering och frånvaro som följd. Att man som idrottsutövare ska behöva ta denna risk i fullkontaktssporter är något man får räkna med, dock att skapa sig en bättre förståelse för problemet och minska antalet hjärnskakningar skulle göra stora framsteg.
Denna rapport ska genom videoanalys i Skillspector ta fram hastigheter i
tacklingsögonblicket därefter genom FEM analyser i programmet LS-Dyna undersöka när hjärnskakning kan förväntas uppkomma.
Videosekvenserna hämtade från YouTube är främst NHL matcher men några övriga sekvenser finns. Inget speciellt urval har gjorts annat än HD-kvalitet och konstaterade hjärnskakningar. En liten referensgrupp utan hjärnskakningar har även inhämtats för att jämföra resultatet.
Totalt utvärderades 31 fall av hjärnskakning och 7 fall där hjärnskakning inte uppkom. Medelvärden för islagshastigheten för de skadade spelarna blev 6.87m/s för attackerande spelare och 4.46m/s för de tacklade spelarna resulterande i en linjär acceleration på 657 m/s2 och en rotationsacceleration på 6883 rad/s2 och ett HIC värde på 731. För de spelare som inte drabbades av hjärnskakning var medelvärdet på hastigheten för de attackerande spelarna 5.83m/s de tacklade spelarna 3.53m/s resulterande i en linjär acceleration på 385m/s2 rotationsacceleration 5474rad/s2 och ett HIC värde på 309.
Abstract
Head injuries such as concussions is a major problem in full contact sports like ice hockey, these injuries usually leads to rehabilitation and absence from sports as result. This risk in full contact sport is something athletes in these sports must expect, however to gain a better understanding of these injuries and reduce the number concussions would be a great progress.
This master thesis will investigate through video analysis in motion tracking software Skillspector velocities in moment of the body check. These velocities will form the basis of the LS-Dyna simulation to conduct when concussion may occur.
The video footage was mostly from NHL but some others are present all collected from YouTube. No special selection was made other than HD quality and some statement where concussions had occurred, a small reference group of non-concussed players was also collected to compare some results.
Table of content
1. Introduction ... 1
2. Video elaboration and analysis ... 6
2.1 Video sample technique ... 6
3. FEM simulation ... 14
4. Results ... 16
5. Discussion and Future work ... 18
6. References ... 20
7. Appendix 1 ... 21
8. Appendix 2 ... 22
1
1. Introduction
Concussion or Mild Traumatic Brain Injury (MTBI) is defined as a complex pathophysiological process affecting the brain, induced by traumatic biomechanical force1.
Several different definition2 of how to categorize MTBI exist the most common are Glasgow coma scale3 (GCS) which measure the level of consciousness by a form where
different levels of reaction are evaluated. Duration of post traumatic amnesia (PTA).
Loss of consciousness (LOC).
These are usually divided in three categories depending on the severity of the brain injury Table 1
GCS PTA LOC
Mild 13-15 <1 hour 0-30min
Moderate 9-12 1 hour-24 hours >30min to <24 hours
Severe 3-8 >1 day >24 hours
Table 1. Classification of TBI
In this report only the mild concussion is represented, but since there is a big secretiveness and all information for each case are not available there is hard to distinguish the true level of the MTBI since the teams cover up the injuries. However when stretcher is brought to the ice it’s possible to suspect a moderate TBI, in this report match 7, Hossa is the one suffered from the longest LOC judging from video footage.
In sport related concussion there are also guidelines whenever a player should return to play after concussions.
Cantu guidelines4 table 2 was the first to truly develop guidelines whenever return to play shall be conducted.
First concussion Second concussion Third concussion
Mild Return to play
when
asymptomatic.
Return to play in two weeks when asymptomatic for one week.
Terminate season.
May return next season.
Moderate Return to play
when
asymptomatic for one week.
Return to play after one month if asymptomatic for one week.
Terminate season.
May return next season.
Severe Return to play
one month after injury if
asymptomatic for one week.
2 Table 2. Cantu guidelines for returning to play.
For Colorado medical society5 the categorization is as table 3.
Grade First concussion Second concussion Third concussion
Mild Return to play
when
asymptomatic for 20 minutes.
Return to play when asymptomatic for one week.
Terminate. May return in three months.
Moderate Return to play when
asymptomatic for one week.
Return to play after asymptomatic for one month.
Terminate season. May return next season. Severe Transport to hospital. Return to play one month after injury if asymptomatic for two weeks. Terminate season. Discourage return
Table 3. Colorado medical society guidelines for returning to play.
The American Academy of Neurology (AAN) 6 which is an improvement of the Colorado medical society categorized according to table 4
First concussion Second concussion Third concussion
Mild Return to play when
asymptomatic for 15 minutes.
Return to play when asymptomatic for one week. If second concussion in same day, activity should be terminated for that day.
Moderate Return to play when asymptomatic for one week.
Return to play after two weeks if asymptomatic.
Severe Return to play if asymptomatic for two weeks.
Return to play if asymptomatic for one month.
Table 4 AAN guidelines for returning to play.
3
There are also suggested values regarding acceleration and HIC value when concussion may occur7.
Probability of MTBI Translational acceleration (g) Rotational acceleration (rad/s2) HIC 25% 62 4800 120 50% 78 6200 240 80% 92 7800 350
Table 5. Threshold for the probability of a concussion to occur. Payne8 defined HIC levels and acceleration according to table 6
Injury level Proposed tolerance level Equivalent acc (g)
HIC 15ms (For 3ms) 1 (no concussion) < 150 < 55 2 (mild concussion) 150-500 55-90 3 (severe concussion) 500-1800 90-150 4 (Life threatening) > 1800 > 150
Table 6. Injury levels according to Payne
Overall just in sport approximately 3.5 million MTBI occur in the United States every year9. Since a part of these concussions has devastating effects of the receiving participator it may be categorized as a major problem. None should be forced to quit his or her favorite sport due to concussions.
In the National Hockey League (NHL) around 5-6% of players suffer from concussion during a full season10. Studies has been conducted that most concussions occur in the first period with majority in the defensive zone, 47% concussions in open ice and 53% in the perimeter. Age around 28 years old and the hitter were on average taller and heavier11. 88% of all concussions involve player to player contact, in some few cases due to fighting. Shoulder to head contact is around 42% of all cases, elbow to head 15%, gloves 5%, in this thesis
shoulder to head is the main point of contact. In 65% of the cases the concussed player was in possession of the puck or had just released it, and in 35% there were no possession of the puck12.
4
a greater way i.e. suspension and fines. The latest call in NHL was to establish a certain rule called Rule 48 that’s to reduce blind side checks which worked out well overall and some of concussions were reduced. Player to player contact is not surprisingly the most common reason for MTBI in ice hockey, many of these players may have a long rehabilitation time ahead and for some players the career is at end. Especially when receiving subsequent concussions the risk dramatically increases as seen in table 2, 3 and 4.
Concussion from ruthless body checks delivered directly to the head in high speed is the ones players need to reduce in order to keep the numbers of MTBI on a low level.
Unfortunate circumstances where players fall into i.e. the board in retrieving the puck is one concussion type that is part of the game hard to get rid of, reducing the number of
concussions to just those cases would be a great progress. The mentality among hockey players especially in youth hockey also need to be improved as 6% says they would consider injury an opponent in order to win a game13.
This master thesis will be using motion analysis build a database with velocities in
concussion body checks and investigate using LS-Dyna prepost14 whether a certain threshold regarding velocity at impact are contributing to concussion.
Motion analysis of video footage was carried out using the free video motion software Skillspector15 which easily can be described as a frame by frame analyzer. The concept of the software is to mark certain point in the video from at least five frames, calibrate the setup and obtain results. The main use of this software is to record athletes in educational use with a measured calibration frame, this thesis will do the motion analysis from video footage collected from internet with no predefined calibration frame.
5
Concussion hits
Match number Attacking player Concussed player League Date Injury level
1 Justin Abdelkader Toni Lydman NHL 4th May 2013 2
2 Abdelkader Vladímir Sobotka World cup 24th May 2014 2
3 Matt Cooke Marc Savard NHL 7th Mar 2010 2
4 David Steckel Sidney Crosby NHL 29th Mar 2012 2
5 Deryk Engelland Justin Abdelkader NHL 14th Dec 2013 2
6 Erik Johansson Erik Haula NHL 12th Oct 2014 2
7 Raffi Torres Marian Hossa NHL 17th Apr 2012 2->3
8 John Scott Loui Eriksson NHL 23rd Oct 2013 2
9 Sebastian Karlsson Magnus Kahnberg SHL 5th Nov 2011 2
10 Scott Stevens Eric Lindros NHL 27th May 2008 2
11 Milan Lucic Ryan Miller NHL 13th Nov 2011 2
12 Willie Mitchell Jonathan Toews NHL 21st Oct 2011 2->3
13 John Moore Dale Weise NHL 27th May 2014 2
14 Douglas Murray Mike Kosta NHL 1st Apr 2014 2
15 Matt Niskanen Jeff Skinner NHL 5th Oct 2014 2
16 Brook Orpik Loui Eriksson NHL 8th Dec 2013 2
17 Richie Regehr Sebastian Karlsson SHL 13th Mar 2014 2
18 Mike Richards David Booth NHL 7th Nov 2009 2->3
19 Mike Rupp T.J Oshie NHL 10th Apr 2014 2
20 Ryan Garbutt Dustin Penner NHL 20th Oct 2013 2
21 Vadimir Shipachyov Pekka Jormakka World Cup 11th May 2014 2->3
22 Andy Sutton Jordan Leopold NHL 16th Apr 2010 2
23 Joe Thornton David Perron NHL 4th Nov 2010 2
24 Ryan White Kent Huskins NHL 15th Apr 2013 2
25 Johan Forsberg Daniel Fernholm SHL 20th Oct 2014 2
26 Brian Mcgrattan Andrew Alberts NHL 29th Dec 2013 2
27 Cody Mcleod Niklas Kronwall NHL 17th Oct 2013 2
28 Douglas Murray Kyle Chipchura NHL 1st Nov 2010 2
29 Chris Neil Johnny Boychuk NHL 25th Feb 2012 2->3
30 Aron Rome Nathan Horton NHL 6th Jun 2011 2->3
31 Noah Welch Max Görtz SHL 6th Apr 2015 2
6
Non concussion hits
1 Radko Gudas Scottie Upshall NHL 9th Oct 2014 Injury
level
2 Alex Killorn Paul Ranger NHL 19th Mar 2014 1
3 Douglas Murray Johan Sundström NHL 10th Apr 2014 1
4 Tyler Myers Dainius Zubrus NHL 16th Nov 2011 1
5 Dion Phaneuf Tuomo Ruutu NHL 24th Jan 2011 1
6 John Scott Mikhail Grabovski NHL 16th Oct 2014 1
7 Brent Seabrook David Backes NHL 19th Apr 2014 1
Table 6 Non concussion hits
2. Video elaboration and analysis
In order to analyze the video some steps were required to perform in order to get the video footage from internet to the computer and further in to the analysis software.
Using the YTD video downloader16 the YouTube clips where easily downloaded to the computer keeping the original quality of the footage.
Converting the YouTube MP4 format to the more generally AVI format using a video
converter17 this software were also used to roughly cut out just the hit from different angles in the video, the main camera and the replay camera respectively used for the 3D analysis. Since Skillspector require raw data AVI files the software Virtual Dub18 was used to further convert the video, this software were also used to extract just the few frames needed for the analyze minimum for the analyze to work out were five but up to twelve were used to get a better accuracy. This software was also used to sample the replay camera to match the first camera where the framerate of 29.97 frames per seconds (fps) was known. All of these converting steps reduced the quality of the video footage why initially HD quality was required.
2.1 Video sample technique
7
carried out by removing every second or third frame depending on the replay cameras frame rate. This was done to have the exact number of frames from both cameras so the sequences would match as seen in figure 1 and 2.
Figure 1 Original video footage of the hit
Figure 2 Replay cameras frame rate sampled to origin number of frames by removing every other frame.
This was done to have the exact same frames so the body check were delivered in the same time i.e. same frame for both sequences, after this step the skillspector motion analysis could be set up.
8 Figure 3a Model wizard for the Skillspector analyze.
10
Defining the segment of players head and players upper arm figure 3c
Figure 3c Defining segments
Set the calibration frame and calibration points seen in figure 3d, in this analysis a 3-D frame was selected so the 3-D velocities could be obtained. Positions seen in table 7 was chosen so ears, chin and forehead could be used as reference since all length around the head is approximately 0,2m with help from sketchup the length could be set.
X Y Z 0 0 0 0 0 0.2 0.2 0 0 0 0.2 0 0 0.2 0.2 0.2 0.2 0
11 Figure 3d Setting the calibration frame and points
12 Figure 4 A schematic view of skillspector
The calibration of the video was the tricky part of the motion analysis since no good fixed reference points was located in every sequence. Several different calibration techniques were tried out but the one most effective was a calibration around the head of targeted player prior to impact. To carry out the distance and location of the points needed to calibrate a 3D editing software19 were used. Since length and width of certain lines i.e. blue and red line in the play was known and visible the calibration frame could be carried out by put them on a coordinate system form in the 3D software and then find the equivalent position in skillspector. Some approximation had to be done though since it was hard to distinguish the exact position of each calibration point.
13
A schematic view of the calibration frame can be seen in picture 5
Picture 5 Calibration in Skillspector
When all this was done Skillspector could calculate the 3D velocities for each corresponding hit and the graph is presented as seen in picture 6. Just the moment of impact was taken into account and the first half of the result graph was considered as measurement noise.
14
3. FEM simulation
For the FEM software analysis in Ls-Dyna two simple upper bodies were studied, they were both equipped with helmets consisting of a plastic shell and an inner layer of foam, the body neck and head consists of several different parts, seen in appendix 2, joint together to build up the entire interface. The attacking player was equipped with an upper body shoulder protection corresponding to the one carried by hockey players, some material properties can be seen in appendix 3. Since the attacker in these sequences usually targets the head as a major point of contact there was no need to model the upper body protection of the targeted player. These players were aligned in a way corresponding prior to the hit for each case. Figure 7. Some approximation regarding impact angels were carried out since angle data from skillspector was not to be extracted. So the angles was found by using the sketchup software and thereafter the dummies were positioned in the best way possible corresponding to the hit.
Figure 7. A view of the Ls-dyna simulation
Velocities from Skillspector analysis were to be set for each case as best way possible, approximation had to be done since 3D velocities from Skillspector was not to be
15
deformation that produces no stress or strain21. Refining the mesh is the right way to preventing this error the other faster way is to apply an hourglass filter, in this case a stiffness based LS-Dyna numbering four were used. This action permits deformation in a greater way due to extra modes and the simulation could be carried out. The drawbacks is of course contribution to the result from these extra modes as they absorb energy however it’s difficult to say exactly what contribution this hourglassing has on the result.
Hourglassing require an energy threshold < ~10% of total internal energy for reliable result which were checked in energy plot and holds in this analysis.
The outcome of the simulation was to investigate linear acceleration rotational acceleration and HIC value, defined as
𝐻𝐼𝐶 = {[ 1 𝑡2 − 𝑡1∫𝑡1 𝑡2 𝑎(𝑡)𝑑𝑡] 2.5 (𝑡2 − 𝑡1)} 𝑚𝑎𝑥
Were t2 is final time and t1 is initial time in an interval where HIC is measured g is head acceleration chosen to maximize the HIC value. This means that both the acceleration and the duration of impact is at great matter for the HIC value. And therefore a greater
acceleration in less time may be accepted regarding HIC value threshold.
For the FEM simulations the outcome where according to the graph in figure 8. These curves values could be extracted and plotted in different software i.e. MATLAB but since just the peak values was read off there was no use for this action.
16
4. Results
The data collected as seen in previously displayed graphs were read of for velocities; acceleration and HIC value respectively and are presented in table 8 the result here are presented as the resultant velocity and acceleration.
There are some results marked with an * where the targeted player was hit towards the boards and needed to be model with this extra addition which was not done in this thesis.
17
Non concussion hits
1 6,11 2,57 429 4010 251 2 4,87 4,23 * * 3 3,67 2,43 * * 4 5,63 5,20 302 6510 442 5 5,19 5,30 305 6950 350 6 5,17 2,73 314 4670 209 7 7,05 2,28 578 5230 296 Average 5.38 3.53 385 5474 309
18
5. Discussion and Future work
Several different situations in professional ice hockey have been evaluated both in velocity and in FEM simulations. The work was found to be far more time consuming than originally thought and not all simulations was controlled in reliability of velocities or by energy plots but samples was collected. Therefore errors may occur but the method used for motion analysis was found to work well and the velocities presented in the motion analysis are reliable from the known velocities of players during a game.
The velocity threshold was not found but there is an overwhelming risk of concussion for a 6 m/s check where the head is picked as a major point of contact resulting in a rotation
acceleration of the head around 6000 rad/s2. For hits more straight forward higher velocities around 8 m/s with translational acceleration 600 m/s2 can be tolerated, these values found by interpolation of corresponding cases. However there is hard to distinguish a threshold with this analysis. The average values were found to be 6.45m/s for attacker velocity, 4.45 m/s for the concussed player. Linear acceleration 657 m/s2, rotational acceleration 6883 rad/s2 and HIC value 731. These values corresponds to around 25% risk of concussion for linear acceleration between 50-80% for rotational acceleration and over 80% in HIC values. This states that rotational acceleration is the most dominant acceleration type and also the one causes most concussions.
For non-injury the values were 5.38 m/s for attacking player, 3.53 m/s for the targeted player, linear acceleration 385m/s2, rotational acceleration 5474 rad/s2 and HIC value 309. The risk in these few samples are less than 25% for linear acceleration, between 25-50% in rotational acceleration and between 50-80% for HIC value according to Newman7. A
fundamental likelihood analysis needs to be performed to draw some statistical conclusions of the results, was not done in this evaluation also additional non injury simulations needed in order to conclude a threshold for concussions.
19
20
6. References
1) Aubry, M., Cantu, R., Dvorak, J., Graf-Baumann, T., Johnston, K., Kelly, J., … Schamasch, P. (2002). Summary and agreement statement of the first international conference on concussion in sport, vienna 2001. The Physician and Sportsmedicine, 30(2), 57–63. doi:10.1136/bjsm.2005.018614
2) http://www.glasgowcomascale.org/
3) Nicholl, J., & LaFrance, W. C. (2009). Neuropsychiatric sequelae of traumatic brain injury. Seminars in Neurology, 29(3), 247–255. doi:10.1055/s-0029-1223878
4) Robert C Cantu R. Return to play guidelines after head injury
5) Miller, J., Wendt, J., & Potter, N. (2011). Implications for concussion assessments and return-to-play standards in intercollegiate football: How are the risks managed? Journal of Applied Sport Management, 3(1), 91–103. Retrieved from
http://quod.lib.umich.edu/j/jsas/6776111.0003.116?rgn=main;view=fulltext
6) Giza, C. C., Kutcher, J. S., Ashwal, S., Barth, J., Getchius, T. S. D., Gioia, G. a,… Zafonte, R. (2013). Summary of evidence-based guideline update: evaluation and management of concussion in sports: report of the Guideline Development Subcommittee of the American Academy of Neurology. Neurology, 80(24), 2250–7. doi:10.1212/WNL.0b013e31828d57dd 7) Newman, J. a, Barr, C., Beusenberg, M., Fournier, E., Shewchenko, N., Welbourne, E., &
Withnall, C. (2000). A New Biomechanical Assessment of Mild Traumatic Brain Injury. Part 2: Results and Conclusions. Proceedings of the 2000 International IRCOBI Conference on the Biomechanics of Impact, 223–233.
8) Payne, A.R. (2001). “Occupant Protection & Egress In Rail Systems,” © MIRA 2001, Project 427519 Version 1.1 http://www.eurailsafe.net/subsites/operas/
9) Langlois, J. a, Rutland-Brown, W., & Wald, M. M. (2006). The epidemiology and impact of traumatic brain injury: a brief overview. The Journal of Head Trauma Rehabilitation, 21(5), 375–378. doi:00001199-200609000-00001 [pii]
10) Sojka, P. (2011). “Sport” and “non-sport” concussions. Cmaj, 183(8), 887–888. doi:10.1503/cmaj.110504
11) Hutchison, M. G., Comper, P., Meeuwisse, W. H., & Echemendia, R. J. (2013). A systematic video analysis of National Hockey League (NHL) concussions, part I: who when where and what. British Journal of Sports Medicine, 1–5. doi:10.1136/bjsports-2013-092235
12) Hutchison, M. G., Comper, P., Meeuwisse, W. H., & Echemendia, R. J. (2013). A systematic video analysis of National Hockey League (NHL) concussions, part II: how concussions occur in the NHL. British Journal of Sports Medicine, 1–5. doi:10.1136/bjsports-2013-092235 13) Marchie, A., & Cusimano, M. D. (2003). Bodychecking and concussions in ice hockey: Should
our youth pay the price? Cmaj, 169(2), 124–128.
14) Ls-dyna prepost www.lstc.com 7374 Las Positas Road, Livermore, CA 94551
Tel: (925) 449-2500 support@lstc.com
15) Video4Coach – SkillSpector, www.video4coach.com, Grubbemøllevej 8 -5700 Svedenborg
Denmark Tel: +45 41606029 Mail: info@video4coach.com
16) YTD Video downloader http://www.ytddownloader.com/ 17) Slysoft Clone DVD http://www.slysoft.com/
18) Virtual dub http://www.virtualdub.org/ 19) sketchup.google.com
20) http://www.putty.org/
21
7.
Appendix 1
22
8. Appendix 2
23
9. Appendix 3
Material properties for some different parts
Rho kg/m3 Yong’s modulus (Pa) Poisson’s Ratio
Head 3546.8 2.05e+011 0.31
Helmet outer 1161.81 1.64e+009 0.45
Helmet inner 40.0 8.0e+006 0
Brain 1040.0 0.50
Neck 2700.0 7.0e+010 0.31
Rigid body, Thorax 9.819e+004 2.05e+011 0.31
Chest protection 40.0 8.0e+006 0
